high-quality surface passivation of silicon using native ... · wafers were cleaned using standard...
TRANSCRIPT
High-quality surface passivation of silicon using native oxide and siliconnitride layersZahidur R. Chowdhury, Kevin Cho, and Nazir P. Kherani Citation: Appl. Phys. Lett. 101, 021601 (2012); doi: 10.1063/1.4733336 View online: http://dx.doi.org/10.1063/1.4733336 View Table of Contents: http://apl.aip.org/resource/1/APPLAB/v101/i2 Published by the American Institute of Physics. Related ArticlesOptimal hydrogenated amorphous silicon/silicon nitride bilayer passivation of n-type crystalline silicon usingresponse surface methodology Appl. Phys. Lett. 101, 171602 (2012) Reduced temperature sensitivity of the polarization properties of hydrogenated InGaAsN V-groove quantumwires Appl. Phys. Lett. 101, 151114 (2012) Tunable electroluminescence from polymer-passivated 3C-SiC quantum dot thin films Appl. Phys. Lett. 101, 123110 (2012) Analysis of sub-stoichiometric hydrogenated silicon oxide films for surface passivation of crystalline silicon solarcells J. Appl. Phys. 112, 054905 (2012) Imaging crystal orientations in multicrystalline silicon wafers via photoluminescence Appl. Phys. Lett. 101, 082102 (2012) Additional information on Appl. Phys. Lett.Journal Homepage: http://apl.aip.org/ Journal Information: http://apl.aip.org/about/about_the_journal Top downloads: http://apl.aip.org/features/most_downloaded Information for Authors: http://apl.aip.org/authors
Downloaded 13 Nov 2012 to 142.1.133.126. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions
High-quality surface passivation of silicon using native oxide and siliconnitride layers
Zahidur R. Chowdhury, Kevin Cho, and Nazir P. Kherania)
Department of Electrical and Computer Engineering, University of Toronto, 10 King’s College Road, Toronto,Ontario M5S 3G4, Canada
(Received 27 January 2012; accepted 20 June 2012; published online 9 July 2012)
We report on the attainment of high quality surface passivation of crystalline silicon using facile
native oxide and plasma enhanced chemical vapour deposition SiNx. Using systematic
measurements of excess carrier density dependent minority carrier lifetime, it is observed that the
inferred interface defect density decreases with increasing native oxide thickness while the
interface charge density remains unchanged with thickness, which ranges from 0.2 A to 10 A. A
surface recombination velocity of 8 cm/s is attained corresponding to a native oxide layer thickness
of �10 A. Similar chemically grown oxide layer followed by SiNx deposition is shown to yield
comparable passivation, indicating practical viability of the passivation scheme. VC 2012 AmericanInstitute of Physics. [http://dx.doi.org/10.1063/1.4733336]
Excellent surface passivation of crystalline silicon using
low temperature processes enables the use of ultra-thin
wafers for photovoltaic (PV) solar cell manufacturing, thus
paving the way for high-efficiency low-cost silicon photovol-
taics. Fabrication of low cost cells using such schemes is via-
ble because of the reduced material costs and low thermal
budget. Low temperature deposition of hydrogenated amor-
phous silicon (aSi:H),1,2 plasma enhanced chemical vapour
deposition (PECVD) of silicon oxide (SiOx),3,4 PECVD sili-
con nitride (SiNx),3–7 and PECVD silicon carbide (SiCx)8
have been reported as potential passivation compounds for
crystalline silicon (cSi) surface.
Out of these passivation schemes, PECVD SiNx has
been studied extensively. Stoichiometric SiNx is deemed
more suitable owing to its lower absorption in the UV region
compared to silicon rich SiNx. Further, stoichiometric SiNx
has better chemical etching selectivity against cSi than sili-
con rich SiNx,6 which is desirable where selective etching of
SiNx is necessary for device fabrication. High values of fixed
positive charge density, Qs, at the cSi-SiNx interface leads to
field effect passivation in the case of nearly stoichiometric
SiNx. Charge densities of more than 1012 cm�2 have been
reported.9,10 But an induced inversion layer due to the high
trapped charge density can lead to parasitic shunting which
reduces the short circuit current.11 A thin layer of thermally
grown SiO2 (Refs. 12–14) or low temperature PECVD
oxide15 layer is deposited between the cSi surface and SiNx
layer in order to reduce or remove the parasitic effect. Excel-
lent passivation using these schemes has been reported, with
surface recombination velocity (SRV) of 10 cm=s (Refs. 10,
12–14) and 6 cm=s (Ref. 10) for the introduced thermal ox-
ide (10 nm thickness) and PECVD oxide (50 nm thickness)
layers, respectively.
This article reports attainment of excellent surface passi-
vation of n-type float zone (FZ) crystalline silicon using con-
trolled facile native oxide growth at room temperature
followed by the deposition of low-temperature PECVD
SiNx. Contrary to the conventional practice of etching, the
native amorphous silicon oxide layer using hydrofluoric acid
prior to SiNx deposition, we show that the presence of a fac-
ile native oxide layer, denoted f SiOx, plays a critical role in
reducing the interface dangling bond defect density, Ns.
Using the dangling bond interface recombination model,16,17
we infer Ns and Qs from excess carrier density (ECD) de-
pendent lifetime measurements, showing interface defect
density as low as 7:2� 109 cm�2 and surface charge density
of 4:3� 1011 cm�2 which correspond to a SRV of less than
8 cm/s at ECD of 1015 cm�3 (7 cm/s at ECD 5� 1014 cm�3).
Double-side polished n-type (100) FZ wafers with resis-
tivity of 1 to 5 X cm were used in this passivation study. The
wafers were cleaned using standard RCA cleaning steps.18
The wafers were subsequently dipped in 5% HF for 2 min to
etch the native oxide and then kept in a cleanroom environ-
ment, with ambient average humidity of 50% and tempera-
ture of 25 �C, for different durations; during this period the
wafers were stored within a typical 25 wafer carrier box. The
duration for the growth of facile native oxide layers prior to
the deposition of SiNx was varied from 20 min to 40
000 min. The 100 mm diameter circular wafers were cut into
four quadrants prior to the deposition of SiNx deposition. Sil-
icon nitride deposition was carried out using the Oxford
PlasmaLab 100 direct rf PECVD system. During a series of
preliminary experiments, it was determined that higher SiNx
deposition temperatures resulted in better passivation, and
accordingly all the SiNx depositions were carried out at
400 �C. While the use of low chamber pressure (200 mTorr)
and high plasma power (100 W) has been reported for SiNx
passivation studies using a similar system,6 our experience
indicated significant non-uniformity in film quality. Instead,
we found that operating at a higher chamber pressure of
1 Torr and lower plasma power of �25 W yielded uniform
and reproducible films. Further, the use of higher chamber
pressure allowed better convective heat transfer from the
substrate holder to the wafer and hence required a lower
a)Also at the Department of Material Science and Engineering, University
of Toronto, Toronto, Ontario M5S 3G4, Canada. Electronic mail:
0003-6951/2012/101(2)/021601/4/$30.00 VC 2012 American Institute of Physics101, 021601-1
APPLIED PHYSICS LETTERS 101, 021601 (2012)
Downloaded 13 Nov 2012 to 142.1.133.126. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions
preheating period compared to that at a lower chamber pres-
sure. Operating at higher plasma power increased the deposi-
tion rate of SiNx albeit with no significant change in the
passivation quality. Accordingly, all SiNx depositions were
carried out at the low plasma power of �25 W and chamber
pressure of 1 Torr.
Precursor gases for silicon nitride depositions consisted
of NH3, which was held constant at a flow rate of 50 sccm,
and 5% silane in nitrogen at flow rates ranging from 200
sccm to 550 sccm. We define the precursor gas ratio (GR) as
the quotient of (SiH4þN2) to NH3 flow rates. The effect of
varying the GR from 4 to 7 was explored for samples where
the facile native oxide growth time (OGT) ranged from
20 min to 2 K min. A small enhancement in passivation was
observed for these relatively short OGTs. On the other hand,
excellent passivation enhancement in quality was observed
for facile native oxide growth times of the order of 40 K min.
For these samples, the GR was varied from 4 to 11. For all
samples, the SiNx deposition time was held constant at
7 min; further, SiNx was deposited on both sides of the wafer
to yield a symmetric set of passivation layers. Thicknesses
and refractive indices of the SiNx films were measured using
SOPRA GES5E spectroscopic ellipsometer.
The ECD dependent effective minority carrier lifetime,
seff , was measured using a Sinton Silicon Lifetime Tester
WCT-120 system. Transient and quasi-steady-state photo-
conductance (QSSPC) methods were used to measure
injection-dependent seff of the sample.19 Spatial distributions
of the lifetime for different samples were measured using
Semilab’s WT-2000 PVN microwave photo conductance
decay (l-PCD) instrument. Oxide thicknesses were meas-
ured using parallel angle resolved x-ray photoelectron spec-
troscopy (PARXPS). Theta Probe from Thermo ScientificVR
was used for the PARXPS measurements.
In fitting the data obtained from SE measurements, we
included a thin native oxide layer between the cSi substrate
and SiNx layer. Silicon nitride layer thickness of �93 nm
thickness was inferred from the SE measurements which cor-
responds to a growth rate of 13.3 nm/min. Refractive index
(at 633 nm wavelength) increased from 1.9 to 2.06 with
increasing silane precursor concentration, i.e., the GR. The
increase in the refractive index is consistent with a transition
from a sub-stoichiometric to a silicon-rich silicon nitride
film.6
seff values at an excess carrier density of 1015 cm�3
(unless stated otherwise) were used to calculate the effective
SRV, Seff ,
Seff ¼ W=ð2 seffÞ; (1)
where W is the thickness of the sample and the bulk lifetime is
assumed to be infinite. Fig. 1 shows the effective SRV values
at an ECD of 1015 cm�3 for samples with a range of native
OGTs as a function of the GR. It is evident from the figure
that the passivation quality for the f SiOx-SiNx layers generally
improves with the facile native oxide growth time and increas-
ing GR. In particular, we find that a reduction in SRV with
OGT is moderate up to OGT of 2 K min, however, significant
improvement in passivation is observed for OGT of 20 K min
and yet a further enhancement for OGT of 40 K min. For the
sample with an OGT of 40 K min, the SRV is reduced further
with an additional increase in the GR from 7 to 11.
The high resolution Si2p data measured by Theta Probe
were used to fit mixed Lorentzian and Gaussian functions
(i.e., Voigt function) with a modified Shirley background. The
relative contributions of oxide and elemental silicon signals
(percentage contributions) were determined from the fitting.
Fig. 2 shows the percentage contributions of oxide and ele-
mental silicon signals varying over a range of incident angles
for different oxide growth times. Theoretical percentage con-
tributions were also calculated for the oxide layers on cSi with
different oxide thicknesses using the Beer-Lambert equation.
Oxide thicknesses were determined based on the best fitting of
the measured and theoretical percentage contributions. The
measured native oxide thicknesses are of same order of
magnitude as reported in Morita et al.’s investigation of native
oxide growth20 on n-type cSi wafer having a doping concen-
tration similar to that used in this study. It is noted that the
XPS measurements carried out by Morita et al.20 used only a
single take-off angle of the photoelectrons.
Fig. 3 shows the passivation quality as a function of
measured native oxide thicknesses for different OGTs. As
FIG. 1. Seff for an ECD of 1015 cm�3 for the facile native oxide (f SiOx)-
SiNx passivation schemes. OGT and the GR for the SiNx deposition were
varied.
FIG. 2. The relative oxide and elemental silicon signals measured using par-
allel angle resolved XPS technique, showing contribution of elemental sili-
con (Si-e) and silicon oxide (Si-o) at different incident angles and oxide
growth time.
021601-2 Chowdhury, Cho, and Kherani Appl. Phys. Lett. 101, 021601 (2012)
Downloaded 13 Nov 2012 to 142.1.133.126. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions
seen from the figure, increase in the native oxide thickness
corresponds to an improvement in the passivation quality.
The inset shows that the facile native oxide thickness pla-
teaus with oxide growth time to a limiting value of 10 A.
Morita et al. reported a limiting oxide thickness of 8 A.20
The surface charge density, QS, and defect density, NS,
are inferred from the ECD dependent Seff and simple closed
form (SCF) dangling bond interface recombination
model.16,17 The measured and SCF fitted Seff are shown in
Fig. 4 for several samples with different native OGT and
GR. The extracted QS and NS values are shown in Figs. 5(a)
and 5(b), respectively. Notwithstanding some scatter in the
data, the average trapped charge density remains essentially
unchanged over the range of OGT studied here as well as the
GR. However, on average, the interfacial dangling bond
defect density decreases with increasing OGT, which corre-
sponds to the observed reduction in SRV. These results indi-
cate the important role of a native oxide layer of less than or
approximately 1 nm in thickness, along with hydrogenated
silicon nitride, in markedly reducing the defect density at the
interface.
A sample with a facile native oxide growth time of
160 K min was also prepared using identical SiNx deposition
conditions stated above and at a GR of 7. Fig. 6(a) shows the
spatial profile of the passivation quality measured using the
l-PCD tool. A corresponding maximum lifetime of 1.7 ms
was measured which is equivalent to Seff of �8 cm=s.
Given the observed importance of the facile oxide layer,
we also explored the potential of a similar chemically grown
oxide playing an equivalent role. Using a 62% nitric acid so-
lution at 60 �C, facile silicon oxide was grown on a quadrant
of a cSi wafer by submerging it in the bath for 60 min. The
spatial distribution of the passivation attained, for a process
that has yet to be optimized, is shown in Fig. 6(b). This
result corresponds to a lifetime of 0.76 ms or a SRV of
18 cm/s as measured using the Sinton lifetime tester. Further
optimization of the chemically grown oxide and PECVD
SiNx passivation is expected to yield results similar to the
native oxide layers grown in ambient atmosphere at room
temperature.
The article presents an excellent passivation scheme for
crystalline silicon surfaces using facile native f SiOx and
PECVD SiNx dual layers. The native f SiOx layer, which is
approximately 1 nm in thickness, can be grown readily using
a number of equivalent techniques at or near ambient tem-
perature, results in very low dangling bond density at the
interface while the field effect passivation due to interface
and trapped charges remains essentially unchanged. The
low-temperature passivation scheme is potentially amenable
FIG. 4. ECD dependent Seff for different GR and OGT. Lines show the fit-
ting using dangling bond interface recombination model in simple closed
form (Refs. 16 and 17).
(a)
(b)
FIG. 5. (a) Interface charge density, QS, and (b) dangling bond defect den-
sity, NS, for different GR as a function of OGT. These values are inferred by
fitting the ECD dependent Seff values using dangling bond interface recom-
bination model.
FIG. 3. Seff for different gas ratios (h GR 4 � GR6 D GR 7) as a function
of the measured facile native oxide thickness measured using PARXPS. The
inset shows the measured (this work and by Morita et al. (Ref. 20)) native
oxide thickness as a function of oxide growth time.
021601-3 Chowdhury, Cho, and Kherani Appl. Phys. Lett. 101, 021601 (2012)
Downloaded 13 Nov 2012 to 142.1.133.126. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions
to high-efficiency ultra-thin silicon photovoltaics and allied
microelectronic and optical devices.
This work was supported by the Ontario Research Foun-
dation - Research Excellence program, Natural Sciences and
Engineering Research Council of Canada, and the University
of Toronto.
1M. Taguchi, M. Taguchi, H. Sakata, and E. Maruyama, Sol. Energy Mater.
Sol. Cells 95, 18 (2011).2U. Rau, N. Jensen, and J. H. Werner, in Proceedings of the 22nd EuropeanPhotovoltaic Solar Energy Conference, Milan, Italy (2007), p. 816.
3C. Leguijt, P. Lolgen, J. A. Eikelboom, P. H. Amesz, R. A. Steeman,
W. C. Sinke, P. M. Sarro, L. A. Verhoef, P. P. Michiels, Z. H. Chen, and
A. Rohatgi, Sol. Energy Mater. Sol. Cells 34, 177 (1994).4Z. Chen, A. Rohatgi, and D. Ruby, in Proceedings of the IEEE 1st WorldConference on Photovoltaic Energy Conversion, Waikoloa, HI (IEEE,
1994), p. 1331.5T. Lauinger, J. Schmidt, A. G. Aberle, and R. Hezel, Appl. Phys. Lett. 68,
1232 (1996).6J. Schmidt and M. Kerr, Sol. Energy Mater. Sol. Cells 65, 585 (2001).7H. Mackel and R. Ludemann, J. Appl. Phys. 92, 2602 (2002).8I. Martin, M. Vetter, A. Orpella, J. Puigdollers, A. Cuevas, and R. Alcu-
billa, Appl. Phys. Lett. 79, 2199 (2001).9W. L. Warren, J. Kanicki, J. Robertson, E. H. Poindexter, and P. J.
McWhorter, J. Appl. Phys. 74, 4034 (1993).10G. Dingemans, M. M. Mandoc, S. Bordihn, M. C. M. van de Sanden, and
W. M. M. Kessels, Appl. Phys. Lett. 98, 222102 (2011).11S. Dauwe, L. Mittelstadt, A. Metz, and R. Hezel, Prog. Photovoltaics 10,
271 (2002).12J. Schmidt, M. Kerr, and A. Cuevas, Semicond. Sci. Technol. 16, 164
(2001).13S. Narasimha and A. Rohatgi, Appl. Phys. Lett. 72, 1872 (1998).14Y. Larionova, V. Mertens, N.-P. Harder, and R. Brendel, Appl. Phys. Lett.
96, 032105 (2010).15M. Hofmann, S. Janz, C. Schmidt, S. Kambor, D. Suwito, N. Kohn, J.
Rentsch, R. Preu, and S. W. Glunz, Sol. Energy Mater. Sol. Cells 93, 1074
(2009).16S. Olibet, E. V.-Sauvain, and C. Ballif, Phys. Rev. B 76, 035326 (2007).17B. Bahardoust, A. Chutinan, K. Leong, A. B. Gougam, D. Yeghikyan, T.
Kosteski, N. P. Kherani, and S. Zukotynski, Phys. Status Solidi A 207, 539
(2010).18J. D. Plummer, M. D. Deal, and P. B. Griffin, Silicon VLSI Technology
Fundamentals, Practice and Modeling (Prentice-Hall, Englewood Cliffs,
NJ, 2000).19R. A. Sinton and A. Cuevas, Appl. Phys. Lett. 69, 2510 (1996).20M. Morita, T. Ohmi, E. Hasegawa, M. Kawakami, and M. Ohwada,
J. Appl. Phys. 68, 1272 (1990).
10
5
0130 255.97 381.93 507.9 633.86 759.83
Lifetime [us]
Surf
ace
[%]
130 759.83
(a)
(b)
457.66 1173.9
1173.91030.7887.4744.16600.91457.66
20151050
Lifetime [us]
Surf
ace
[%]
FIG. 6. Spatial profile of the lifetime measured using l-PCD tool. (a) Native
oxide-SiNx passivation where facile native oxide was grown for 4 months.
(b) Chemically grown oxide and SiNx passivation. HNO3 solution was used
for the oxide growth.
021601-4 Chowdhury, Cho, and Kherani Appl. Phys. Lett. 101, 021601 (2012)
Downloaded 13 Nov 2012 to 142.1.133.126. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions